3.1.2

Python Kernel

• Fixed an installation issue when there are multiple versions of Python

ONNX Kernel

• Added ORT DML Runtimes for generation speedup (40 sec -> 10 sec)
• Fixed GPU detect for AMD Vega

ONNX Features

• LoRA support
• LoCONS and LoHA partial support

Other

• LoRA subdirectories support
• Multiply LoRA using (auto111 style)

The release was soon due to the correction of some problems, so some Preview features that have not yet been adapted to the user's UI setup got here. Exactly:

• Models preview image:
• • just add logo.png/logo.jpg file to model dir
• LoRA tokens
• • make models\lora(lora_name).txt file and write tokens. UI will upload them

PR

• add experimental support for LoHA networks by @ssube in #5
• handle 3x3 kernels in LoHA better by @ssube in #6

New Contributors

• @ssube made their first contribution in #5

Full Changelog: 3.1.1...3.1.2

3.1.1

What is new

• ONNX converter fixed

XUI 3.1

What is new

• ControlNet: Support depth/hed/seg/mlsd/normalmap/canny/scribble (CUDA/ONNX)
• PyTorch 2.0: fixes for ONNX converter
• Host: install/uninstall packages
• • py -install onnx
• • py -uninstall pil

XUI 3.0

General

• Improved application stability
• The generator operation mode is implemented without unloading the model
• • And other...

Other Changed

• Update SHARK driver requirement by @Amblyopius in #3

New Contributors

• @Amblyopius made their first contribution in #3

Full Changelog: 3.0...3.0.1

XUI 3.0 (Preview)

• Stable improved
• Added waifu upscaler
• Added ControlNet support + Pose (CUDA)
• Added LoRA support
• Added Pix2Pix for CUDA/ONNX
• Reading png meta info
• • And other

Full Changelog: 2.3...3.0

XUI 2.3

• Fixed GFPGAN install
• Fixed DeepDanbooru working with low VRAM
• Added new user notifications for errors/process

XUI 2.2

• Fixed setup script
• Fixed UI
• Improved UX

XUI 2.1

• Fixed CUDA venv setup
• Fixed UI bugs
• Added DeepDanbooru (tokens extract)

And others

New UI and CUDA support

• Added CUDA and ONNX support.
• Added converters
• New UI
• Auto installing

Bugs:

• All models use diffus. Need convert to shark for Diffusers(CPU/CUDA).

1.0

Shark API + Vulkan upscalers.
See readme for install